[Ffmpeg-devel-irc] ffmpeg.log.20141024

burek burek021 at gmail.com
Sat Oct 25 02:05:01 CEST 2014


[00:04] <t4nk363> hi
[00:04] <Tatou> hi
[00:04] <t4nk363> any reason why vp8 does not like odd numbers where scaling/croping?
[00:05] <t4nk363> scale=max(180\,a*135):max(135\,180/a),crop=180:135 -> gives 180x134
[00:05] <kevmitch> that's not unique to vp8. most codecs are like that
[00:05] <c_14> It's actually not the codec, it's the pixel format.
[00:05] <t4nk363> or more like scale=180:135, crop=180:135:0:0
[00:05] <c_14> yuv420p only likes even numbers
[00:05] <kevmitch> for one the chroma (color) data is stored at half the resolution
[00:05] <t4nk363> aha, so what do you suggest
[00:05] <t4nk363> ?
[00:05] <c_14> make it even?
[00:05] <kevmitch> use even sizes
[00:05] <t4nk363> yeah i know that, but what if i need odd
[00:06] <c_14> use yuv444p
[00:06] <t4nk363> k, will try, thanks for the help
[00:16] <t4nk363> Incompatible pixel format 'yuv444p' for codec 'libvpx', auto-selecting format 'yuv420p'
[00:17] <t4nk363> even if i use libvpx-v9, same result
[00:17] <c_14> Hmm, seems vp8 only supports yuv420 and yuva420
[00:18] <c_14> Why exactly do you need an odd size?
[00:18] <t4nk363> thats the requirements i got :)
[00:19] <c_14> Can it be one bigger or one smaller?
[00:19] <c_14> Can you pad it so it's one bigger or one smaller?
[00:19] <c_14> A one pixel black border shouldn't be that bad.
[00:19] <c_14> Though a one pixel stretch usually isn't noticable either.
[00:20] <t4nk363> yeah, probably i will go with 134
[00:21] <kevmitch> http://ffmpeg-users.933282.n4.nabble.com/VP9-with-yuv444p-subsampling-td4663040.html
[00:22] <t4nk363> but look at this
[00:22] <t4nk363>            "width": 180,             "height": 135,             "has_b_frames": 0,             "sample_aspect_ratio": "1:1",             "display_aspect_ratio": "4:3",             "pix_fmt": "yuv420p",
[00:23] <kevmitch> a fairly recent git commit 04b0dda853198980a14158809e560ea9ca9d7e33  looks to have hadded support for more than just 420
[00:24] <kevmitch> wait, that's just decoding
[00:24] <t4nk363> why with ogv the size can be 180x135..
[00:25] <kevmitch> and requires --enable-experimental  --enable-vp9-high config options
[00:25] <kevmitch> ogv is a container. what codec? vpx, theora?
[00:26] <t4nk363> theora
[00:26] <kevmitch> who knows, maybe it stores a pad pixel and truncates it for display
[00:27] <c_14> I actually think several codecs do something similar.
[01:03] <t4nk363> strange, even padding doesnt work
[01:38] <kevmitch> t4nk363: define doesn't work
[05:40] <t4nk511> Hello. I'm trying to submit a patch to ffmpeg-devel at ffmpeg.org but for some reason it doesn't seem to be making it to the mailing list (for example when I check here: https://lists.ffmpeg.org/pipermail/ffmpeg-devel/2014-October/date.html).
[05:41] <t4nk511> The commands I run are `git format-patch -1 --subject-prefix='FFmpeg-devel] [PATCH'`
[05:41] <t4nk511> That outputs a .patch file, which I then try to submit with: `git send-email blah.patch`
[05:42] <t4nk511> which asks for confirmation, then I say yes, and it then says Result: OK
[05:43] <t4nk511> oh, just realized I'm in the user-support channel. Re-posting in the ffmpeg-devel channel...
[07:17] <ruby_on_tails> i want to add a 8px border on the top and bottom of a video, i tried this: ffmpeg -i df1.mp4 -vf pad=1360:8:0:0:black df1-border.mp4
[07:17] <ruby_on_tails>  but it doesnt work it says [pad @ 0x185f480] Input area 0:0:1360:752 not within the padded area 0:0:1360:8 or zero-sized Error opening filters!
[07:34] <relaxed> ruby_on_tails: what's the input frame size?
[07:34] <ruby_on_tails> 1360*752
[07:34] <ruby_on_tails> i want to make it 1360*768 by adding 8px black borders on top and bottom
[07:39] <relaxed> try -vf pad=1360:768:8:0:black
[09:36] <termos> I am now setting rtmp_buffer to 15 seconds, but it does not seem to introduce a 15 seconds delay. Is this value honored in FFmpeg?
[09:46] <pikovit> hey guys, any ideas about this problem from a couple of days ago... http://pastebin.com/eRUqSSTT
[16:26] <BlackDream> Hello, i'm trying to use PROXY with FFmpeg i noticed there is an httpproxy protocol in the available protocols
[16:26] <BlackDream> However i can't make it work
[16:26] <BlackDream> ffmpeg -i "httpproxy://proxy_ip:proxy_port/http://real_stream"
[16:26] <BlackDream> it is working like that? i didnt find any documentation
[17:37] <florolf> hi
[17:39] <florolf> i was just reading through mpegtsenc.c (http://git.videolan.org/?p=ffmpeg.git;a=blob;f=libavformat/mpegtsenc.c;h=d73b75fc9240000115b6b314728e79e58f356598;hb=HEAD#l232)
[17:39] <florolf> shouldn't the block calculating the program info length (246-248) be placed *after* the loop?
[17:40] <florolf> the way it's written right now will always write a 0 there
[17:46] <c_14> I'm not sure of the specifics myself, but you'd probably be better off asking in #ffmpeg-devel or on the mailing list[s].
[17:52] <rcombs> florolf: huh, interesting
[17:52] <rcombs> c_14's right; probably best to ask on the ML
[17:52] <florolf> okay
[18:01] <florolf> rcombs: ffmpeg-devel? not sure if this counts as a "bug report". i'm also not submitting a patch (yet).
[18:08] <c_14> Personally, I'd probably send to -devel
[18:09] <florolf> okay
[19:49] <chuckp> what would I need to do this on a raspberry pi? http://www.raspberrypi.org/forums/viewtopic.php?t=45893
[19:49] <chuckp> I'm not wanting to do the varnish
[19:52] <OlafLostViking> Hi! I am using unpaper which seems to use ffmpeg. Since the last update to unpaper 6 I am getting the error "No codec provided to avcodec_open2()" as shown in pastebin.com/MTmZQWVy . Is there a known incompatibility or do I need to install additional packages provided by ffmpeg (that unpaper isn't finding at the moment)? (using Archlinux)
[19:59] <OlafLostViking> ltrace pastebin.com/axgDy6eq , strace pastebin.com/WgM4khMV
[20:24] <danomite> Are there any tutorials or examples on how to stream an IP camera?
[20:26] <edoardo> danomite: i successfully streamed a v4l device to another pc but it depends on the protocol of the camera
[20:27] <danomite> its an rtsp wrapped h264 video, no audio needed.
[20:48] <esing> Hi, how can I convert mp3 to aax?
[20:55] <kepstin-laptop> esing: you mean aac?
[20:56] <kepstin-laptop> first, do you have a really good reason to do so? lossy transcoding should be avoided whenever possible :)
[20:56] <kepstin-laptop> esing: and once you've decided that, read https://trac.ffmpeg.org/wiki/Encode/AAC if you still want to do it for some reason.
[20:57] <esing> kepstin-laptop: I mean aax, the audible format for amazon audiobooks. It's also mp3, but with the aax container
[20:58] <kepstin-laptop> huh, that's confusing. since x and c are beside eachother in qwerty layout, it looks just like a typo :)
[20:59] <esing> Hm I think I actually need the aa format, since aax seems to be the container for mpeg4 http://en.wikipedia.org/wiki/Audible.com#Quality
[21:01] <kepstin-laptop> esing: hmm, that just means the 'mp4' container, the codec is still mp3
[21:01] <kepstin-laptop> so you could just do something like ffmpeg -i file.mp3 -c:a copy file.mp4 and then rename the output, maybe
[21:02] <kepstin-laptop> might not accept it without some special meta info in the file tho.
[21:03] <esing> Mhm I see
[21:26] <sanooj> hey guys. I'm having trouble feeding my logitech c920 to a youtube live stream.  youtube claims it's not receiving any data from the encoder
[21:28] <sanooj> but iftop shows bits certainly moving that direction at about the expected speed
[21:29] <sanooj> I'm running ./t/bin/ffmpeg -f v4l2 -vcodec h264 -i /dev/video0 -copyinkf -vcodec copy  -f flv 'rtmp://x.rtmp.youtube.com/live2/<stream id here>'
[21:31] <sanooj> if I send it to a file instead of an rtmp server it works great, so I'm a little confused.
[22:06] <Sevanoo> Hi
[22:06] <sanooj> Hi Sevanoo.
[22:06] <Sevanoo> I have some problems with ffprobe
[22:07] <Sevanoo> sometimes when i check my streams it tell me thats not working. is there an way to make it check like 15 sec
[22:07] <Sevanoo> retry
[22:08] <sanooj> no idea. I'm a total newb. just being friendly. :)
[22:09] <Sevanoo> hehe no problem :) thanks to be friendly:)
[22:11] <BlackDream> Sevanoo add this in the beggining of the command  -probesize 100M -analyzeduration 250M
[22:11] <BlackDream> ffprobe will get more accurate details
[22:12] <Sevanoo> oke thanks i will try
[22:23] <sanooj> can I upgrade the profile of an h264 stream from Contrained Baseline to Main without recompressing it it?
[22:25] <Mavrik> sanooj, uh, no.
[22:26] <sanooj> I'm thinking youtube might be dropping my live stream on the floor because it's constrained baseline instead of Main 4.1
[22:26] <Mavrik> That's like asking if you can turn english into german without rewriting the words ;)
[22:26] <Mavrik> I'm pretty sure YouTube can handle constrained baseline, since that's what most devices are recording nowdays :)
[22:27] <sanooj> do you happen to know if they require an audio stream?  I don't have one
[22:27] <kepstin-laptop> constrained baseline is a subset of main; im pretty sure; anything that can handle decoding main should be able to decode constrained baseline
[22:28] <sanooj> how can I check the key frame frequency of my stream?
[22:29] <sanooj> mostly I'm stumped because I don't know what's wrong with my camera's stream that's causing it to claim it's not getting any data
[22:31] <Mavrik> hmm, it's possible that they need an audio stream yeah
[22:32] <sanooj> hey hang on. is my stream's url even correct?  rtmp://x.rtmp.youtube.com/live2/<stream id here>
[22:32] <sanooj> they give the stream id and the url up to .../live2 separately. I took a wild guess and concatenated like the above
[22:35] <kepstin-laptop> that looks like it should be correct
[22:35] <kepstin-laptop> if you just want to try adding some silence, throw in a -filter_complex 'aevalsrc=c=2:exprs=0|0' to get some silence, and add appropriate audio codec options
[22:39] <sanooj> thanks kepstin-laptop. I'll try that once I'm convinced of my url. I'm seeing other kinds of urls on the internet ending in .../videolive?<PARAMETERS HERE>, but they don't go into what the parameters are :(
[22:40] <kepstin-laptop> I saw one note of a user who had a url like yours working, but they noted that it didn't work until they set the audio bitrate to some minimum value
[22:40] <kepstin-laptop> dunno
[22:41] <sanooj> aha. lemme try the audio then.
[22:41] <sanooj> youtube, y u no have error messages?
[23:16] <sanooj> apparently the audio thing was it
[23:16] <sanooj> once I added some audio youtube is no longer dropping it on the floor. now it's complaining I'm not sending keyframes often enough
[23:17] <sanooj> which is an improvement
[23:33] <edoardo> danomite: ffmpeg -i rtsp://@192.168.1.1:12345 -acodec copy -vcodec copy c:/abc.mp4
[23:46] <sanooj> so... it looks like I should be able to tune the key frame period because there's some experimental code in a libv4l2-hw-code library from 2012 that prints a bunch of encoding parameters
[23:47] <sanooj> including dwFrameInterval and wIFramePeriod
[23:47] <sanooj> but how do I set those parameters in ffmpeg?
[23:49] <c_14> You just want more keyframes?
[23:50] <c_14> set -g
[23:50] <c_14> (i think)
[23:51] <sanooj> my webcam is doing the encoding
[23:51] <sanooj> I'll try that.
[23:54] <sanooj> it says: Codec AVOption g (set the group of picture (GOP) size) specified for input file #1 (/dev/video0) is not a decoding option.
[23:55] <sanooj> unfortunately v4l2-ctl doesn't list any of the encoding parameters as controls
[23:58] <sanooj> I'm pretty sure I shouldn't need to be hacking ioctls into ffmpeg to get this working
[23:58] <c_14> Are you using -codec copy?
[23:58] <sanooj> yes
[23:59] <c_14> mhm
[23:59] <sanooj> because the webcam is doing the encoding.
[23:59] <c_14> In that case you'll have to convince your webcam to give you more keyframes.
[00:00] --- Sat Oct 25 2014


More information about the Ffmpeg-devel-irc mailing list